Android Taint Flow Analysis for App Sets Will Klieber*, Lori Flynn, Amar Bhosale , Limin Jia, and Lujo Bauer Carnegie Mellon University *presenting
Motivation Detect malicious apps that leak sensitive data. E.g., leak contacts list to marketing company. “All or nothing” permission model. Apps can collude to leak data. Evades precise detection if only analyzed individually. We build upon FlowDroid . FlowDroid alone handles only intra-component flows. We extend it to handle inter-app flows. 2
Introduction: Android Android apps have four types of components : Activities (our focus) Services Content providers Broadcast receivers Intents are messages to components. Explicit or implicit designation of recipient Components declare intent filters to receive implicit intents. Matched based on properties of intents, e.g.: Action string (e.g., “ android.intent.action.VIEW ”) Data MIME type (e.g., “ image/png ”) 3
Introduction Taint Analysis tracks the flow of sensitive data. Can be static analysis or dynamic analysis. Our analysis is static. We build upon existing Android static analyses: FlowDroid [1]: finds intra-component information flow Epicc [2]: identifies intent specifications [1] S. Arzt et al., “FlowDroid: Precise Context, Flow, Field, Object-sensitive and Lifecycle-aware Taint Analysis for Android Apps”. PLDI , 2014 . [2] D. Octeau et al., “Effective inter-component communication mapping in Android with Epicc: An essential step towards holistic security analysis”. USENIX Security , 2013 . 4
Our Contribution We developed a static analyzer called “ DidFail ” (“Droid Intent Data Flow Analysis for Information Leakage”). Finds flows of sensitive data across app boundaries. Source code and binaries available at: (or google “DidFail SOAP”) http://www.cert.org/secure-coding/tools/didfail.cfm Two-phase analysis: 1. Analyze each app in isolation. 2. Use the result of Phase-1 analysis to determine inter-app flows. We tested our analyzer on two sets of apps. 5
Terminology Definition. A source is an external resource (external to the app, not necessarily external to the phone) from which data is read. Definition. A sink is an external resource to which data is written. For example, Sources : Device ID, contacts, photos, current location, etc. Sinks : Internet, outbound text messages, file system, etc. 6
Motivating Example App SendSMS.apk sends an intent (a message) to Echoer.apk , which sends a result back. SendSMS.apk Echoer.apk Device ID (Source) getIntent() startActivityForResult() onActivityResult() setResult() Text Message (Sink) SendSMS.apk tries to launder the taint through Echoer.apk. Existing static analysis tools cannot precisely detect such inter-app data flows. 7
Analysis Design Phase 1 : Each app analyzed once, in isolation. FlowDroid: Finds tainted dataflow from sources to sinks. Received intents are considered sources. Sent intent are considered sinks. Epicc: Determines properties of intents. Each intent-sending call site is labelled with a unique intent ID . Phase 2 : Analyze a set of apps: For each intent sent by a component, determine which components can receive the intent. Generate & solve taint flow equations. 8
Running Example src 1 Three components: C 1 , C 2 , C 3 . I 1 C1 = SendSMS C 1 sink 1 C2 = Echoer C 2 I 3 src 3 C3 is similar to C1 C 3 sink 3 • sink 1 is tainted with only src 1 . • sink 3 is tainted with only src 3 . 9
Running Example src 1 I 1 C 1 sink 1 C 2 I 3 src 3 C 3 sink 3 Notation: 10
Running Example src 1 I 1 C 1 sink 1 C 2 I 3 src 3 C 3 sink 3 Notation: 11
Running Example src 1 I 1 C 1 sink 1 C 2 I 3 src 3 C 3 sink 3 Final Sink Taints: • T(sink 1 ) = {src 1 } Notation: • T(sink 3 ) = {src 3 } 12
Phase-1 Flow Equations Analyze each component separately. Phase 1 Flow Equations : src 1 C 1 sink 1 C 2 src 3 C 3 sink 3 Notation • An asterisk (“ ∗ ”) indicates an unknown component. 13
src 1 Phase-2 Flow Equations I 1 C 1 sink 1 C 2 Instantiate Phase-1 equations for all I 3 src 3 possible sender/receiver pairs. C 3 sink 3 Phase 1 Flow Equations : Phase 2 Flow Equations: Notation 14
src 1 Phase-2 Taint Equations I 1 C 1 sink 1 For each flow equation “src → sink”, C 2 I 3 src 3 generate taint equation “T(src) ⊆ T(sink)”. C 3 sink 3 Phase 2 Flow Equations : Phase 2 Taint Equations: Notation If s is a non-intent source, then T( s ) = { s }. 15
Phase 1 Epicc Original APK TransformAPK FlowDroid (modified) Extract manifest 16
Implementation: Phase 1 APK Transformer Assigns unique Intent ID to each call site of intent-sending methods. Enables matching intents from the output of FlowDroid and Epicc Uses Soot to read APK, modify code (in Jimple), and write new APK. Problem: Epicc is closed-source. How to make it emit Intent IDs? Solution (hack): Add putExtra call with Intent ID. Phase 1 Epicc Original APK TransformAPK FlowDroid (modified) Extract manifest 17
Implementation: Phase 1 FlowDroid Modifications: Extract intent IDs inserted by APK Transformer, and include in output. When sink is an intent, identify the sending component. In base.startActivity , assume base is the sending component. (Soundness?) For deterministic output: Sort the final list of flows. Phase 1 Epicc Original APK TransformAPK FlowDroid (modified) Extract manifest 18
Implementation: Phase 2 Phase 2 Take the Phase 1 output. Generate and solve the data-flow equations. Output: 1. Directed graph indicating information flow between sources, intents, intent results, and sinks. 2. Taintedness of each sink. 19
Testing DidFail analyzer: App Set 1 SendSMS.apk Reads device ID, passes through Echoer, and leaks it via SMS Echoer.apk Echoes the data received via an intent WriteFile.apk Reads physical location (from GPS), passes through Echoer, and writes it to a file 20
Testing DidFail analyzer: App Set 2 (DroidBench) Int3 = I( IntentSink2.apk, IntentSource1.apk, id3 ) Graph generated using GraphViz. Int4 = I( IntentSource1.apk, IntentSink1.apk, id4 ) Res8 = R(Int4) Src15 = getDeviceId Snk13 = Log.i Some taint flows : 21
Limitations Unsoundness Inherited from FlowDroid/Epicc Native code, reflection, etc. Shared static fields Implicit flows Currently, only activity intents Bugs Imprecision Inherited from FlowDroid/Epicc DidFail doesn’t consider permissions when matching intents All intents received by a component are conflated together as a single source 22
Use of Two-Phase Approach in App Stores We envision that the two-phase analysis can be used as follows: An app store runs the phase-1 analysis for each app it has. When the user wants to download a new app, the store runs the phase-2 analysis and indicates new flows. Fast response to user. 23
DidFail vs IccTA IccTA was developed (at roughly the same time as DidFail) by: Li Li, Alexandre Bartel, Jacques Klein, Yves Le Traon (Luxembourg); Steven Arzt, Siegfried Rasthofer, Eric Bodden (EC SPRIDE); Damien Octeau, Patrick McDaniel (Penn State). IccTA uses a one-phase analysis IccTA is more precise than DidFail’s two-phase analysis. Two-phase DidFail analysis allows fast 2nd-phase computation. Future collaboration between IccTA and DidFail teams? 24
Conclusion We introduced a new analysis that integrates and enhances existing Android app static analyses. Demonstrated feasibility by implementing a prototype and testing it. Two-phase analysis can be used by app store to provide fast response. Future work: Implicit flows Static fields Distinguish different received intents Other data channels (file system, non-activity intents) Etc. 25
Thank You
Recommend
More recommend